The dashboard tells the owner where the AI stack is working. The ROI calculation tells the owner what each tool in the stack is worth in dollars. Not "Avoca seems helpful" — "Avoca produced $487,000 of attributable revenue last twelve months on a $36,000 subscription, payback in 27 days, ROI 13.5x." Not "Rilla feels like a good investment" — "Rilla lifted close rate 14 points across four Comfort Advisors at $14,200 average ticket on 42 monthly leads, attributable revenue $312,000, subscription $19,200, ROI 16.3x." Every owner conversation with a coach, peer group, banker, franchisor, or PE partner about AI in 2026 ends at the same question: show me the math. This lesson is the math for the four tools that move the biggest numbers — Avoca (missed-call recovery), Rilla (close-rate coaching), ServiceTitan Dispatch Pro (revenue-per-truck yield), Hatch (stale-lead reactivation) — plus the discipline that separates real ROI from vibes ROI. Do the math, not the vibes.

Why ROI per Tool, Not ROI per Stack

The temptation when defending an AI stack is to roll the whole thing up: "we spend $66K/year on AI tools and got $1.4M in incremental revenue, so it's clearly working." That defense holds in informal conversations and fails in any rigorous review. The coach asks which tool to double down on. The peer group asks which tool to pilot. The PE partner asks which tool to kill. The franchisor asks which tool produces the platform-wide lift before they mandate it across 80 sites. None of those questions can be answered from a stack-level number; all of them require per-tool attribution.

Per-tool ROI also exposes the freeloaders. Most shops by month nine of an AI deployment have one tool producing 60-70% of the attributable lift, two tools producing 20-30%, and one or two tools producing single-digit ROI that the owner forgets about until the renewal email lands. Rolling everything into stack-level ROI hides the freeloader; per-tool ROI surfaces it. The Q4 renewal cycle is where freeloaders get killed and budget gets reallocated to the high-ROI tools — and that decision requires per-tool numbers that survive a coach's challenge.

The discipline of per-tool ROI also produces vendor-management leverage. The owner walks into the Avoca renewal conversation in March 2027 with "you produced $487K of attributable revenue last year on $36K — I want to renew at the same rate or roll to a 24-month deal with a 5% lift cap." The Avoca rep cannot easily counter the math because the math is integrated into the FSM data. The owner without per-tool ROI walks into the same conversation with "Avoca seems helpful, we'd like to renew" and gets the standard 12-18% annual lift the vendor's sales team scripts. Per-tool ROI is also negotiation leverage at the contract table.

Avoca ROI — The Missed-Call Recovery Math

Avoca's ROI is the cleanest in the trades AI stack because the workflow it serves — missed-call recovery — produces the most directly attributable revenue. The formula is straightforward. Avoca attributable revenue = recovered booked calls × completion rate × average ticket × gross margin. Each term comes from a system the shop already runs.

Recovered booked calls is the count of inbound calls that Avoca answered and booked that would have leaked without it — after-hours calls, in-hours overflow during CSR busy bands, abandoned-during-hold calls. The number lives in Avoca's missed-call dashboard joined to the FSM appointment record. At a typical 7-truck shop running 200 inbound calls/week with a 22% pre-Avoca missed-call rate, the recoverable pool is ~44 calls/week. Avoca's HL Bowman case study and similar 2026 deployments show 75-85% capture of the recoverable pool — roughly 33-37 booked calls/week, or ~1,700 calls/year of recovery.

Completion rate is the percentage of booked calls that show and complete (do not no-show or cancel). The shop's overall completion rate post-deployment runs 90-94%; Avoca-recovered calls run slightly lower at 86-90% because some of the recovered after-hours bookings are speculative homeowner inquiries. Practical completion rate for Avoca attribution: 88%. Average ticket at the shop's blended mix: $620 (service + repair + replacement) — but Avoca-recovered calls skew slightly higher because they include after-hours emergency work at $750-$1,100 average. Practical average ticket for Avoca attribution: $680. Gross margin at typical shop economics: 38%.

Putting it together at the 7-truck shop: 1,700 recovered calls/year × 88% completion × $680 average ticket × 38% gross margin = ~$386,000 of attributable gross margin contribution. Subtract subscription ($36K/year for a 7-truck Avoca tier in 2026) = $350K net contribution. Payback: $36K / ($386K / 12) = 1.1 months. ROI: ($386K - $36K) / $36K = 9.7x in first year, accelerating in years 2-3 as the recovered customers become repeat customers and produce membership signups that compound the original recovery into recurring revenue.

The audit discipline. Quarterly: pull 30 random Avoca-attributed bookings; validate each against the FSM appointment record to confirm the call would have leaked without Avoca (counterfactual rigor — homeowner says "I'd have called a competitor if you didn't answer" rather than "I'd have called back tomorrow"). Target: 70%+ high-counterfactual. Below 50% means Avoca attribution is inflated and the ROI claim does not survive a PE board challenge.

Rilla ROI — The Close-Rate Lift Math

Rilla's ROI sits inside the Comfort Advisor team's economics and is structurally different from Avoca's. Rilla doesn't produce calls; it lifts the close rate on calls the team already runs. The formula: Rilla attributable revenue = close-rate lift × leads × average ticket × gross margin × advisor count.

Close-rate lift is the delta between pre-Rilla and post-Rilla close rate on replacement leads. Rilla's published 2026 data shows 18% close-rate lift across home-services deployments — typically expressed as moving from 38-42% baseline close to 52-58% post-deployment. The lift is fragile and lives entirely inside the daily 15-minute coaching huddle the L3 Ch4 workflow runs; without the huddle, Rilla is a recording. With the huddle, the lift holds across 12+ months of deployment. Leads per advisor per month: 40-50 replacement-eligible homeowner visits at a typical 4-advisor team. Average ticket on replacement: $14,200 residential heat-pump / furnace / panel-upgrade mix (HVAC). Gross margin on replacement: 32%. Advisor count: 4.

Putting it together at the 4-advisor team with 14-point close-rate lift (high end of the band): 0.14 lift × 42 leads/advisor/month × $14,200 ticket × 32% margin × 4 advisors × 12 months = ~$361,000 of attributable gross margin/year. Subtract subscription ($300/seat/month × 4 advisors × 12 = $14,400/year, plus $200/month manager seat = $16,800 total) = ~$344K net contribution. Payback: $16,800 / ($361K / 12) = 0.56 months (17 days). ROI: ($361K - $16.8K) / $16.8K = 20.5x.

The audit discipline. Quarterly: pull 30 random Rilla-attributed closed deals; validate the AI-scored ride-along card was actually reviewed in a coaching huddle within 72 hours of the kitchen-table close; validate the advisor's pre-Rilla baseline close rate was measured rather than assumed; validate the post-Rilla close rate uses the same denominator definition (replacement-eligible homeowner visits, not all sales-advisor touches). Target: 70%+ of closed deals have documented huddle coaching trace. Below 50% means the lift is being claimed without the operational discipline that produces it; ROI claim is fragile.

Dispatch Pro ROI — The $/Truck Delta Math

ServiceTitan Dispatch Pro's ROI sits in the dispatch yield lift — the delta between pre-Dispatch-Pro and post-Dispatch-Pro RPT (revenue per truck per day). The formula: Dispatch Pro attributable revenue = RPT lift × trucks × days × gross margin.

RPT lift at typical 2026 shops running Dispatch Pro for 60+ days with override discipline: $87-$120/truck/day on the residential service fleet. The number comes from comparing the trailing-60-day pre-deployment RPT against the trailing-60-day post-deployment RPT, controlling for seasonal mix shift (compare summer to summer, winter to winter, not summer to spring). At HL Bowman's published case and similar 2026 deployments, the 12-18% RPT lift band holds — the high end of which translates to $190-$270/truck/day on a base of $1,800 service RPT. The practical attribution: $145/truck/day at a typical shop with calibrated override discipline. Trucks: 7 service trucks at the example shop. Days: 250 operating days/year. Gross margin on service revenue: 42%.

Putting it together at the 7-truck shop with $145/truck/day RPT lift: $145 × 7 trucks × 250 days × 42% margin = ~$107,000 of attributable gross margin/year. Subtract subscription ($4,200/year for Dispatch Pro add-on at 7-truck tier) = $103K net contribution. Payback: $4,200 / ($107K / 12) = 0.47 months (14 days). ROI: ($107K - $4.2K) / $4.2K = 24.5x.

Dispatch Pro produces the highest first-year ROI of any tool in the stack because the subscription is small and the lift sits on top of an existing revenue base that the dispatcher was already producing — the AI is finding the marginal $87-$120/truck/day that the dispatcher couldn't see while juggling 19 calls on the board.

The audit discipline. Quarterly: pull 30 random Dispatch Pro-recommended dispatches; validate the override rate is in the 5-15% target band (below 5% means the dispatcher is rubber-stamping; above 25% means the AI is mis-calibrated); validate the documented override reasons (comp plan, install crew, recall risk, customer request) make operational sense; pull pre-deployment baseline RPT against original ServiceTitan records to confirm baseline integrity. Target: 70%+ of overrides documented; 70%+ of Dispatch Pro-recommended dispatches accepted. Below 50% acceptance means override discipline is broken and the lift is being eroded.

Hatch ROI — The Stale-Lead Revival Math

Hatch's ROI is the most variable in the stack because it depends entirely on the size and freshness of the dormant lead pile and the segment design discipline the marketing manager runs. The formula: Hatch attributable revenue = dormant lead count × reactivation rate × close rate × average ticket × gross margin.

Dormant lead count is the size of the unworked-lead pile in the FSM by the time Hatch is deployed — typically 2,000-5,000 leads at a $5M shop with 24+ months of operating history (every homeowner who got a quote and didn't close, every after-hours booking that no-showed, every GLSA lead that picked up the phone and disappeared). Practical number at the example shop: 3,200 dormant leads. Reactivation rate: Hatch's published 2026 case studies show 30-45% reactivation on stale leads through AI-drafted text and email sequences. Practical attribution: 35%. Close rate on reactivated leads: 20-30% (lower than fresh leads because the homeowner has cooled but the segment-design discipline tunes which segments to target). Practical attribution: 24%. Average ticket on reactivated leads: $7,800 (mix of service-tuneup re-engagement and replacement-quote follow-up). Gross margin: 35% blended.

Putting it together at the example shop: 3,200 dormant leads × 35% reactivation × 24% close × $7,800 ticket × 35% margin = ~$733,000 of attributable gross margin one-time over the first 12-18 months. The number is one-time because Hatch is working a finite pile; subsequent years produce smaller numbers as the pile gets worked through. Subtract subscription ($4,800/year at the $400/month tier) = $728K net contribution year one. Payback: $4,800 / ($733K / 12) = 0.08 months (3 days). ROI year one: 152x. Steady-state ROI years 2-3: 8-12x on the freshly-cooled-lead pipeline that the shop produces month over month.

Hatch is the highest first-year ROI tool because it converts a previously-unworked asset (the dormant lead pile sitting in the FSM) into closed revenue. The ROI compresses in years 2-3 because the pile is finite; the steady-state ROI lives in the dripping cohort of newly-dormant leads (homeowners who quoted last quarter and went cold) that requires the marketing manager's segment-design discipline to convert.

The audit discipline. Quarterly: pull 30 random Hatch-attributed closed deals; validate the homeowner was actually on the Hatch nurture sequence (not organically re-engaged via NiceJob review-prompt or direct outreach); validate the close happened after the Hatch sequence touchpoint within a defensible attribution window (typically 60 days); validate the dormant-lead-pool count was measured pre-deployment rather than estimated. Target: 70%+ high-attribution rigor. Below 50% means Hatch attribution is inflated and the ROI claim erodes.

The Stack-Level Roll-Up and the Quarterly Tool Review

The per-tool calculations add up to the stack-level number the owner defends at the quarterly board review, the franchisor QBR, the bank refinance conversation, or the PE partner check-in. At the 7-truck example shop running Avoca + Rilla + Dispatch Pro + Hatch in 2026:

Avoca: $386K attributable gross margin, $36K subscription, 9.7x ROI. Rilla: $361K, $16.8K, 20.5x. Dispatch Pro: $107K, $4.2K, 24.5x. Hatch: $733K (year one), $4.8K, 152x year one, 8-12x steady state. Total year-one attributable margin: ~$1.587M. Total year-one AI subscription cost: ~$61.8K. Total year-one stack ROI: 25.7x. Payback period: 23 days. The numbers defend the AI stack at any board, coach, peer-group, or PE conversation with documented per-tool detail.

The quarterly tool review is where this ROI roll-up gets pressure-tested. Three questions per tool: (1) Is the attributable revenue claim defensible against double-counting, baseline integrity, counterfactual rigor? Pull the quarterly attribution audit memo. (2) Is the workflow that produces the lift still running — daily Avoca CSR review, daily Rilla coaching huddle, override discipline on Dispatch Pro, segment design on Hatch? Pull the L3 audit cadence. (3) Is the tool worth renewing at current price against alternative vendors or in-software AI substitutes (ServiceTitan Voice substituting for Avoca, CallRail CI substituting for Rilla on the CSR floor, Sera profit-aware scheduling substituting for Dispatch Pro)? Run the bake-off rubric.

The quarterly tool review produces three decisions per tool: renew at current price, renegotiate (lower price or longer term), or kill and replace. The 90-day-budget review takes the decisions into the next quarter's plan. The pattern across 2026 trades shops running this discipline: 60-70% of tools renew, 20-30% renegotiate, 5-10% get killed and replaced — and the killed-and-replaced category is where the freeloaders surface that the stack-level ROI was hiding.

The Disciplines That Separate Real ROI from Vibes ROI

Five disciplines protect the ROI calculations from being challenged at a coach review, PE board, or franchisor QBR. One: baseline integrity. Pre-deployment baseline numbers must be measured from original FSM and CallRail records, not estimated. The HL Bowman case is defensible because the $350 baseline cost-per-conversion was measured pre-Avoca; the post-deployment $215 number is measured post-Avoca; the 39% reduction is the documented delta. Assumed baselines produce ROI claims that any rigorous review defeats.

Two: counterfactual rigor. Each attributed booking, close, dispatch, or reactivation must answer the question "would this revenue have happened without the AI tool?" Pull 30 samples per tool quarterly. Tag each as high-counterfactual ("homeowner said they'd have called a competitor") or low-counterfactual ("homeowner said they'd have called back themselves"). Healthy: 70%+ high. Below 50%: ROI claim is inflated by attribution to revenue that would have happened anyway.

Three: double-counting audit. No booking can be attributed to two tools simultaneously. If a call books on Avoca's after-hours capture AND the resulting close gets attributed to Rilla's coaching lift AND the dispatch gets attributed to Dispatch Pro's yield AND the customer was re-engaged via Hatch, the same dollar of revenue is being claimed four times. The discipline: each attribution rule has primary ownership; secondary contributions are documented but not counted in the headline ROI. Avoca owns booking attribution; Rilla owns close-rate attribution at the kitchen-table; Dispatch Pro owns dispatch-yield attribution; Hatch owns dormant-lead-pile attribution.

Four: workflow trace. Every claimed lift must trace to a documented workflow running daily. Avoca's lift traces to the 4 p.m. CSR review. Rilla's lift traces to the 15-minute morning coaching huddle. Dispatch Pro's lift traces to the override-reason logging at 5-15% override rate. Hatch's lift traces to the marketing manager's monthly segment design review. Without workflow traces, the ROI claim rests on the vendor's marketing copy and does not survive any operational review.

Five: the trailing-12-month window. ROI claims report against trailing 12 months, not best quarter or launch-month numbers. Vendors push case studies that report best-30-day numbers; the operationally legible ROI uses trailing 12 months because that smooths seasonal variance, workflow drift, and one-time effects. The trailing-12-month ROI is what the owner defends; the launch-month number is the marketing copy.

Key Takeaways

• Per-tool ROI exposes freeloaders and produces vendor-management leverage. Stack-level numbers ($66K spend, $1.4M lift) defend in casual conversation; per-tool numbers defend in PE board reviews, franchisor QBRs, and renewal negotiations.
• Avoca ROI math: recovered calls × completion × average ticket × margin. At 7-truck shop: 1,700 recovered calls/yr × 88% completion × $680 × 38% margin = $386K attributable gross margin on $36K subscription. 9.7x first-year ROI, payback 1.1 months.
• Rilla ROI math: close-rate lift × leads × ticket × margin × advisors. At 4-advisor team with 14-point lift on $14,200 replacement at 42 leads/advisor/month at 32% margin: $361K attributable on $16.8K subscription. 20.5x ROI, payback 17 days.
• Dispatch Pro ROI math: RPT lift × trucks × days × margin. At 7-truck shop with $145/truck/day lift × 250 days × 42% margin = $107K attributable on $4.2K subscription. 24.5x ROI, payback 14 days — highest first-year ROI in the stack because subscription is small and lift sits on existing revenue base.
• Hatch ROI math: dormant leads × reactivation × close × ticket × margin. At 3,200-lead dormant pile, 35% reactivation, 24% close, $7,800 ticket, 35% margin: $733K attributable (one-time over 12-18 months) on $4,800 subscription. 152x year-one ROI; 8-12x steady-state on freshly-cooled-lead drip.
• Stack roll-up at example 7-truck shop: $1.587M attributable margin, $61.8K subscription, 25.7x ROI, payback 23 days. Numbers defend at any board, coach, peer-group, PE conversation with documented per-tool detail.
• Quarterly tool review questions: Is attributable revenue claim defensible (baseline integrity, counterfactual rigor, double-counting)? Is the workflow that produces lift still running daily? Is tool worth renewing vs. alternative vendors or in-software substitutes? Decisions: renew, renegotiate, or kill-and-replace.
• Five disciplines separate real ROI from vibes ROI: baseline integrity (measured not estimated), counterfactual rigor (70%+ high-counterfactual on quarterly sample), double-counting audit (each booking attributed to one primary tool), workflow trace (documented daily cadence underneath each lift), trailing-12-month window (not launch-month vendor case study numbers).
• Per-tool ROI is the renewal-negotiation artifact: "you produced $487K of attributable revenue last year on $36K — renew at current rate or roll to 24-month deal with 5% lift cap." Vendor cannot easily counter the math because the math is integrated into the FSM data. Owners without per-tool ROI accept the standard 12-18% annual lift the vendor's sales team scripts.